Advanced Methods for Image Resolution Enhancement Using Machine Learning

 Advanced Methods for Image Resolution Enhancement Using Machine Learning

Ms.Neeta P.Kulkarni (npkulkarni@coe.sveri.ac.in), Assistant Professor, E&TC Dept., SVERI’s College of Engineering, Pandharpur

Can we imagine our life without images? Imagine a situation where some astronomical research is going on, some space images are taken for analysis which is going to lead to some invention & if these images are not of good resolution, can you get a proper result? Consider another example. A doctor has to do analysis of x-ray of patient. If the image is not with good resolution it cannot produce proper result. Images with high resolution are important for many applications. High resolution images produce better result. Image processing (Image Resolution Enhancement)has many applications in the field of Medical, Satellite Image Processing, Industrial applications. Image resolution enhancement includes improving quality of image by increasing number of pixels, so that the image is more suitable for any applications. It is illustrated in diagram.

a) High Resolution Image  b) Moderate Resolution Image    c) Low  Resolution image 

The traditional methods involved in image processing:

  1. Interpolation (Nearest Neighbor, Bilinear, Bicubic).
  2. Different Transform Methods (Complex Wavelet Transform).
  3. Discrete Wavelet Transforms.
  4. Combination of Discrete Wavelet Transforms &Stationary Wavelet Transforms.
  5. Combination of Interpolation, Discrete Wavelet Transforms & Stationary Wavelet Transforms.

In all these traditional methods, different codes were implemented by using MATLAB, for Interpolation, Wavelet Transforms by using different filters.But all these methods are facing drawbacks like blurring, edge loss ,complexity increases as the interpolation factor increases.

Deep learning for image processing:

To overcome the above drawbacks deep learning is used for image resolution enhancement. It is latest platform used in image processing. In deep learning the methods used for image resolution enhancement involve neural network, deep convolutional networks, Generative Adversarial Networks It gives the better results, more PSNR ratio, less blurring as compared to traditional methods.

Image Super resolution:

Image Super resolution is the technology which allows increasing the resolution of image using deep learning. It is a process of up scaling& improving the details within image. The problem this deep learning method is trying to solve is that the traditional algorithm based up scaling methods lack fine detail and cannot remove defects and compression artifacts. For humans who carry out these tasks manually it is a very slow and painstaking process.

The benefits are gaining a higher quality image from one where that never existed or has been lost; this could be beneficial in many areas or even life saving on medical applications.

To accomplish this mathematical function takes the low resolution image that lacks details and hallucinates the details and features onto it. In doing so the function finds detail potentially never recorded by the original camera.

This mathematical function is known as the model and the up scaled image is the model’s prediction.

Generative Adversarial Networks GANs for Super resolution:

Most deep learning based super resolution model is trained using Generative Adversarial Networks (GANs).

One of the limitations of GANs is that they are effectively a lazy approach as their loss function, the critic, is trained as part of the process and not specifically engineered for this purpose. This could be one of the reasons many models are only good at super resolution and not image repair.

Universal application

Many deep learning super resolution methods can’t be applied universally to all types of image and almost all have their weaknesses. For example a model trained for the super resolution of animals may not be good for the super resolution of human faces.

The model trained with the methods detailed in this article seemed to perform well across varied dataset including human features, indicating a universal model that is effective at upscaling on any category of image may be possible.

Image repair and in painting:

Models that are trained for super resolution should also be useful for repairing defects in a image (jpeg compression, tears, folds and other damage) as the model has a concept of what certain features should look like, for example materials, fur or even an eye.

Image in painting is the process of retouching an image to remove unwanted elements in the image, such as a wire fence. For training it is common to cut out sections of the image and train the model to replace the missing parts based on prior knowledge of what should be there. Image in painting is a usually a very slow process when carried out manually by a skilled human.

References:

https://ieeexplore.ieee.org/

http://towardsdatascience.com

http://www.researchgate.net

Comments